Actin dynamics and competition for myosin monomer govern the sequential amplification of myosin filaments

Abstract

The cellular mechanisms governing non-muscle myosin II (NM2) filament assembly are largely unknown. Using EGFP-NM2A knock-in fibroblasts and multiple super-resolution imaging modalities, we characterized and quantified the sequential amplification of NM2 filaments within lamellae, wherein filaments emanating from single nucleation events continuously partition, forming filament clusters that populate large-scale actomyosin structures deeper in the cell. Individual partitioning events coincide spatially and temporally with the movements of diverging actin fibres, suppression of which inhibits partitioning. These and other data indicate that NM2A filaments are partitioned by the dynamic movements of actin fibres to which they are bound. Finally, we showed that partition frequency and filament growth rate in the lamella depend on MLCK, and that MLCK is competing with centrally active ROCK for a limiting pool of monomer with which to drive lamellar filament assembly. Together, our results provide new insights into the mechanism and spatio-temporal regulation of NM2 filament assembly in cells.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.

from$8.99

All prices are NET prices.

Figure 1: NM2A filament clusters are derived from single nucleation events.
Figure 2: NM2A filament nucleation and expansion produce lamellar networks and ventral stress fibres.
Figure 3: Partitioning of NM2A filaments.
Figure 4: Actin dynamics drive partitioning.
Figure 5: RLC kinase activity regulates NM2A filament partitioning and cluster growth.
Figure 6: NM2A expression level regulates partitioning and filament growth.

References

  1. 1

    Heissler, S. M. & Manstein, D. J. Nonmuscle myosin-2: mix and match. Cell Mol. Life Sci. 70, 1–21 (2013).

    CAS  Article  Google Scholar 

  2. 2

    Ma, X. & Adelstein, R. S. The role of vertebrate nonmuscle Myosin II in development and human disease. Bioarchitecture 4, 88–102 (2014).

    Article  Google Scholar 

  3. 3

    Billington, N., Wang, A., Mao, J., Adelstein, R. S. & Sellers, J. R. Characterization of three full-length human nonmuscle myosin II paralogs. J. Biol. Chem. 288, 33398–33410 (2013).

    CAS  Article  Google Scholar 

  4. 4

    Niederman, R. & Pollard, T. D. Human platelet myosin. II. In vitro assembly and structure of myosin filaments. J. Cell Biol. 67, 72–92 (1975).

    CAS  Article  Google Scholar 

  5. 5

    Kendrick-Jones, J., Smith, R. C., Craig, R. & Citi, S. Polymerization of vertebrate non-muscle and smooth muscle myosins. J. Mol. Biol. 198, 241–252 (1987).

    CAS  Article  Google Scholar 

  6. 6

    Pollard, T. D. Structure and polymerization of Acanthamoeba myosin-II filaments. J. Cell Biol. 95, 816–825 (1982).

    CAS  Article  Google Scholar 

  7. 7

    Adelstein, R. S. & Conti, M. A. Phosphorylation of platelet myosin increases actin-activated myosin ATPase activity. Nature 256, 597–598 (1975).

    CAS  Article  Google Scholar 

  8. 8

    Daniel, J. L. & Adelstein, R. S. Isolation and properties of platelet myosin light chain kinase. Biochemistry 15, 2370–2377 (1976).

    CAS  Article  Google Scholar 

  9. 9

    Amano, M. et al. Phosphorylation and activation of myosin by Rho-associated kinase (Rho-kinase). J. Biol. Chem. 271, 20246–20249 (1996).

    CAS  Article  Google Scholar 

  10. 10

    Craig, R., Smith, R. & Kendrick-Jones, J. Light-chain phosphorylation controls the conformation of vertebrate non-muscle and smooth muscle myosin molecules. Nature 302, 436–439 (1983).

    CAS  Article  Google Scholar 

  11. 11

    Scholey, J. M., Taylor, K. A. & Kendrick-Jones, J. Regulation of non-muscle myosin assembly by calmodulin-dependent light chain kinase. Nature 287, 233–235 (1980).

    CAS  Article  Google Scholar 

  12. 12

    Trybus, K. M., Huiatt, T. W. & Lowey, S. A bent monomeric conformation of myosin from smooth muscle. Proc. Natl Acad. Sci. USA 79, 6151–6155 (1982).

    CAS  Article  Google Scholar 

  13. 13

    Suzuki, H., Onishi, H., Takahashi, K. & Watanabe, S. Structure and function of chicken gizzard myosin. J. Biochem. 84, 1529–1542 (1978).

    CAS  Article  Google Scholar 

  14. 14

    Verkhovsky, A. B., Svitkina, T. M. & Borisy, G. G. Myosin II filament assemblies in the active lamella of fibroblasts: their morphogenesis and role in the formation of actin filament bundles. J. Cell Biol. 131, 989–1002 (1995).

    CAS  Article  Google Scholar 

  15. 15

    Fenix, A. M. et al. Expansion and concatenation of non-muscle myosin IIA filaments drive cellular contractile system formation during interphase and mitosis. Mol. Biol. Cell 27, 1465–1478 (2016).

    CAS  Article  Google Scholar 

  16. 16

    Shiqiong, H. et al. Long-range self-organization of cytoskeletal myosin II filament stacks. Nat. Cell Biol. http://dx.doi.org/10.1038/ncb3466 (2017).

  17. 17

    Beach, J. R. et al. Nonmuscle myosin II isoforms coassemble in living cells. Curr. Biol. 24, 1160–1166 (2014).

    CAS  Article  Google Scholar 

  18. 18

    Li, D. et al. Extended-resolution structured illumination imaging of endocytic and cytoskeletal dynamics. Science 349, aab3500 (2015).

    Article  Google Scholar 

  19. 19

    Suarez, C. et al. Profilin regulates F-actin network homeostasis by favoring formin over Arp2/3 complex. Dev. Cell 32, 43–53 (2015).

    CAS  Article  Google Scholar 

  20. 20

    Rotty, J. D. et al. Profilin-1 serves as a gatekeeper for actin assembly by Arp2/3-dependent and -independent pathways. Dev. Cell 32, 54–67 (2015).

    CAS  Article  Google Scholar 

  21. 21

    Suarez, C. & Kovar, D. R. Internetwork competition for monomers governs actin cytoskeleton organization. Nat. Rev. Mol. Cell Biol. 17, 799–810 (2016).

    CAS  Article  Google Scholar 

  22. 22

    Murugesan, S. et al. Formin-generated actomyosin arcs propel T cell receptor microcluster movement at the immune synapse. J. Cell Biol. 215, 383–399 (2016).

    CAS  Article  Google Scholar 

  23. 23

    Zhang, Y. et al. Mouse models of MYH9-related disease: mutations in nonmuscle myosin II-A. Blood 119, 238–250 (2012).

    CAS  Article  Google Scholar 

  24. 24

    Shutova, M. S., Spessott, W. A., Giraudo, C. G. & Svitkina, T. Endogenous species of mammalian nonmuscle myosin IIA and IIB include activated monomers and heteropolymers. Curr. Biol. 24, 1958–1968 (2014).

    CAS  Article  Google Scholar 

  25. 25

    Guo, M. et al. Probing the stochastic, motor-driven properties of the cytoplasm using force spectrum microscopy. Cell 158, 822–832 (2014).

    CAS  Article  Google Scholar 

  26. 26

    Comrie, W. A., Babich, A. & Burkhardt, J. K. F-actin flow drives affinity maturation and spatial organization of LFA-1 at the immunological synapse. J. Cell Biol. 208, 475–491 (2015).

    CAS  Article  Google Scholar 

  27. 27

    Chew, T. L., Wolf, W. A., Gallagher, P. J., Matsumura, F. & Chisholm, R. L. A fluorescent resonant energy transfer-based biosensor reveals transient and regional myosin light chain kinase activation in lamella and cleavage furrows. J. Cell Biol. 156, 543–553 (2002).

    CAS  Article  Google Scholar 

  28. 28

    Totsukawa, G. et al. Distinct roles of ROCK (Rho-kinase) and MLCK in spatial regulation of MLC phosphorylation for assembly of stress fibers and focal adhesions in 3T3 fibroblasts. J. Cell Biol. 150, 797–806 (2000).

    CAS  Article  Google Scholar 

  29. 29

    Newell-Litwa, K. A. et al. ROCK1 and 2 differentially regulate actomyosin organization to drive cell and synaptic polarity. J. Cell Biol. 210, 225–242 (2015).

    CAS  Article  Google Scholar 

  30. 30

    Decker, B. & Kellermayer, M. S. Periodically arranged interactions within the myosin filament backbone revealed by mechanical unzipping. J. Mol. Biol. 377, 307–310 (2008).

    CAS  Article  Google Scholar 

  31. 31

    Sinard, J. H., Stafford, W. F. & Pollard, T. D. The mechanism of assembly of Acanthamoeba myosin-II minifilaments: minifilaments assemble by three successive dimerization steps. J. Cell Biol. 109, 1537–1547 (1989).

    CAS  Article  Google Scholar 

  32. 32

    Sinard, J. H. & Pollard, T. D. Acanthamoeba myosin-II minifilaments assemble on a millisecond time scale with rate constants greater than those expected for a diffusion limited reaction. J. Biol. Chem. 265, 3654–3660 (1990).

    CAS  PubMed  Google Scholar 

  33. 33

    Luo, T. et al. Understanding the cooperative interaction between myosin II and actin cross-linkers mediated by actin filaments during mechanosensation. Biophys. J. 102, 238–247 (2012).

    CAS  Article  Google Scholar 

  34. 34

    Mahajan, R. K., Vaughan, K. T., Johns, J. A. & Pardee, J. D. Actin filaments mediate Dictyostelium myosin assembly in vitro. Proc. Natl Acad. Sci. USA 86, 6161–6165 (1989).

    CAS  Article  Google Scholar 

  35. 35

    Applegate, D. & Pardee, J. D. Actin-facilitated assembly of smooth muscle myosin induces formation of actomyosin fibrils. J. Cell Biol. 117, 1223–1230 (1992).

    CAS  Article  Google Scholar 

  36. 36

    Mahajan, R. K. & Pardee, J. D. Assembly mechanism of Dictyostelium myosin II: regulation by K+, Mg2+, and actin filaments. Biochemistry 35, 15504–15514 (1996).

    CAS  Article  Google Scholar 

  37. 37

    Ren, Y. et al. Mechanosensing through cooperative interactions between myosin II and the actin crosslinker cortexillin I. Curr. Biol. 19, 1421–1428 (2009).

    CAS  Article  Google Scholar 

  38. 38

    Bockholt, S. M. & Burridge, K. An examination of focal adhesion formation and tyrosine phosphorylation in fibroblasts isolated from src-, fyn-, and yes- mice. Cell Adhes. Commun. 3, 91–100 (1995).

    CAS  Article  Google Scholar 

Download references

Acknowledgements

The authors thank X. Wu, J. Sellers, S. Heissler, N. Billington, A. Pasapera, M. Baird, V. Swaminathan, L. Greene, E. Eisenberg, A. Doyle, T. Egelhoff, L. Lavis, M. Gastinger, the NHLBI Flow Cytometry Core, and GE Deltavision for reagents, help with data acquisition and analysis, critical reading of the manuscript, and helpful discussions.

Author information

Affiliations

Authors

Contributions

J.R.B. and J.A.H. conceived the study. J.R.B. and K.S.B. performed the experiments and analysed the data, with the help of L.S., D.L. and E.B. for TIRF-SIM imaging, K.R., Y.Z., M.A.C. and R.S.A. for studies on MEFs, and Z.S. and N.M.R. for studies with flies. J.R.B. and J.A.H. wrote the manuscript with contributions from the other authors.

Corresponding authors

Correspondence to Jordan R. Beach or John A. Hammer.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Integrated supplementary information

Supplementary Figure 1 Larger fields of view.

Images shown are larger fields of view for the cropped and zoomed images (yellow boxes) shown in Fig. 3 and Fig. 4. The leading edge in A is identified with the cyan line. The leading edge in B is outside the field of view towards the bottom of the image. Pseudocolors, proteins labeled, and original figure location are indicated in each panel. Scale bars are 3 μm.

Supplementary Figure 2 Conservation of partitioning.

(A) COS-7 cells expressing EGFP-NM2B imaged with TIRF-SIM. Magenta box in A corresponds to insets A1A13. Cyan line indicates leading edge. Scale bars represent 2 μm in A and 300 nm in A1. Time in min:sec. A partitioning event is apparent in A10A13. See corresponding Supplementary Video 6. (B) Attempts were made to image the dynamics of endogenous NM2B in the lamella of MEFs isolated from EGFP-NM2B knock-in mice. These attempts were unsuccessful because the signal was too dim (NM2B is expressed at a significantly lower level than NM2A in these cells1), and because this dim signal was present primarily in regions behind the lamella (NM2B typically displays a more central/posterior localization than NM2A1,2). By expressing NM2B-halo-JF549 (magenta) in EGFP-NM2A-MEFs (cyan) and imaging with Airyscan, we were able, however, to observe NM2A filaments containing varying levels of NM2B near the leading edge (B1B3). Upon reaching the rear portion of the lamella where clusters have formed and partitioning is still readily occurring, NM2B appears ubiquitously present in filaments (B-4). We can conclude, therefore, that NM2B undergoes filament partitioning independently of NM2A (because COS-7 cells do not express NM2A3), and that heterotypic filaments containing NM2A and NM2B partition together, even though there is not a substantial amount of NM2B in the anterior lamella of most polarized cells. White boxes in B correspond to insets B1B4. Scale bars represent 2 μm in B and 300 nm in B1 and B4. See corresponding Supplementary Video 8. (CF) Prepupae expressing GFP-squash were punctured and cells that bled were allowed to adhere to the coverslip and imaged with TIRF-SIM. Magenta box in C corresponds to insets D1-D3, which show three different time points where two different partitioning events occurred in the same region. Cyan and yellow boxes in D correspond to E1E8 and F1F8 in lower rows, where individual partition events were tracked over time. Time points indicated in min:sec in bottom right corners. Scale bars represent 3 μm in C and 300 nm in D and E. See corresponding Supplementary Video 7.

Supplementary Figure 3 Methods of analysis.

(AD) Overhead view (A) and orthogonal view (B) of individual EGFP-NM2A (magenta) head group (group ‘a’ from partition in Fig. 4C) following 3D rendering with Imaris software. The surface-surface contact area (yellow) between actin and the NM2A punctum was determined by a novel surface-surface contact algorithm provided by Imaris. Insets C1C3 and D1D3 correspond to xy slices ‘C’ and ‘D’ in (B) and provide examples of how the algorithm works. Note the analysis is done for the entire volume but we show just two slices as examples. Briefly, a surface shell 1 voxel thick (white) was identified that covered the surface of the EGFP-NM2A punctum (magenta) (C2 and D2). Any surface shell voxel that overlapped with the actin surface (as seen in Fig. 4C2 and 4C3) was designated as the surface-surface contact area, shown as yellow in C3 and D3. The number of yellow surface-surface contact voxels divided by the total voxels in the surface shell provided the contact percentage (see Fig. 4D for data). The XYZ coordinate plane in the lower right corner of A and B indicates scale (300 nm in each direction) and orientation. Scale bar in C1 represents 300 nm. (EH) Representative examples of a control actin regions used for surface-surface contact analysis. (E) The raw actin channel from an EGFP-NM2A MEF (collected as described in Fig. 4C). Colored boxes FH in (E) correspond to insets to the right showing the Imaris 3D rendering of the actin networks used for analysis. A mid-partition event was identified in box F (yellow) and the corresponding actin designated ‘Experimental’. Two neighboring actin regions were identified with similar actin densities (G and H, magenta and cyan) and designated ‘Control 1’ and ‘Control 2’. After the surface-surface contact analysis was performed between the NM2A puncta and the experimental actin, the NM2A channel was added to each actin control region and the surface-surface contact analysis was performed again. Scale bars represent 3 μm in E and 500 nm in F. (I) EGFP-NM2A MEFs were imaged with TIRF-SIM every 5 s. Raw data shown in top row. Maxima were identified using the Find Maxima program in ImageJ, are shown in the middle row, and were used to quantitate partitioning rate. The bottom row shows a merge between the raw EGFP-NM2A image (cyan) and the Maxima (magenta) determination. (J) EGFP-NM2A MEFs were imaged with Airyscan every 2 secs. Raw data shown in top row and raw data with mask overlays (magenta) are shown in bottom row. Masks were identified using auto-thresholding in ImageJ. The integrated intensity inside the mask regions were used to quantitate filament cluster growth rate. For I and J, time is indicated in min:sec. Scale bars represent 500 nm. See corresponding Supplementary Video 16.

Supplementary Figure 4 ROCK inhibition and washout.

An EGFP-NM2A MEF (cyan) expressing mApple-F-Tractin (magenta) was imaged with Airyscan before (A1 and A1 ), during (A2A4 and A2 A4 ), and following (A5A7 and A5 A7 ) treatment with Y27632. Scale bars in A1 and A1 represent 5 μm and 3 μm, respectively.

Supplementary Figure 5 Models of NM2 filament partitioning.

Following initial nucleation and growth (far left, cyan), partitioning of filaments might occur through two mechanisms: templated-nucleation (A) or filament splitting (B). In the templated-nucleation model, a mature or maturing NM2 filament (cyan) on an actin fiber (grey) would template the nucleation of a nascent daughter NM2 filament (magenta). This new filament would then grow through the addition of new monomer. The interaction between the initial NM2 filament and the new NM2 filament could be driven by tail-tail interactions and/or head-head interactions4,5,6,7, and could involve additional regulatory components. Recruitment of NM2 monomers and/or filaments through mechano-accumulative properties could also contribute to our observations8,9,10. Dynamic movement of actin fibers (orange double arrow) would then separate these filaments from one another, creating new opportunities for each filament to template additional nascent filaments. In the filament splitting model, similar to one proposed by Fenix et al. 11, a single NM2 filament bound to two actin fibers would be split as the two fibers move away from one another, creating two new daughter filaments that would then grow through the addition of new monomer and then repeat the process.

Supplementary Figure 6 A feedforward mechanism for macro-scale actomyosin network formation.

Actin-dependent NM2 filament partitioning may function in a feedforward fashion to couple the transition for actin from dynamic to stable with the transition for myosin from nascent filaments to large-scale actomyosin structures. Specifically, in the lamellipodium (LP) and anterior lamella where actin is very dynamic, NM2 filament partitioning will be strongly promoted. As more partitioning occurs, the increased number of NM2 filaments will drive the bundling of actin required for stress fiber and transverse arc formation deeper in the lamella. As this bundling proceeds, and higher-order actomyosin structures are assembled in central and posterior regions, the concomitant reduction in actin dynamics will suppress partitioning, which is no longer required. Given that the ultimate goal is to create dense actomyosin structures that power large scale cellular outputs, having a system wherein both NM2 and actin reciprocally promote the amount and organization of the other would provide a self-organizing nature to actomyosin network formation.

Supplementary information

Supplementary Information

Supplementary Information (PDF 14186 kb)

NM2A filaments clusters are derived from single nucleation events.

EGFP-NM2A MEF sampled every 5 s with TIRF-SIM. Playback rate is 30 frames per second. Time in min:sec. Magenta box on the left indicates inset on the right. Scale bars are 2 μm on the left and 300 nm on the right. (AVI 60471 kb)

NM2A filament nucleation and expansion produce lamellar networks and ventral stress fibers.

EGFP-NM2A-MEF, imaged with Airyscan, sampled with 0.5 μm steps every 30 s. Video displays maximum z-projection pseudocolored by depth (color depth scale in top right). Playback rate is 30 frames per second. Time in min:sec. Scale bar represents 3 μm. (AVI 33367 kb)

NM2A clusters feed dorsal networks.

Zoom in from Video 2, region marked in Fig. 2a inset box B. Sampled with 0.5 μm steps every 30 s (color depth scale top right). Playback rate is 30 frames per second. Time in min:sec. Scale bar represents 2 μm. (AVI 3260 kb)

NM2A clusters feed ventral networks.

Zoom in from Video 2, region marked in Fig. 2a inset box C. Sampled with 0.5 μm steps every 30 s (color depth scale top right). Playback rate is 30 frames per second. Time in min:sec. Scale bar represents 2 μm. (AVI 1130 kb)

TIRF-SIM reveals filament partitioning.

EGFP-NM2A MEF sampled every 2 s with TIRF-SIM. Playback rate is 15 frames per second. Time in min:sec. Scale bar represents 300 nm. (AVI 3157 kb)

NM2B partitioning.

COS-7 cell expressing EGFP-NM2B sampled every 10 s with TIRF-SIM. Magenta box in the larger view indicates the zoomed in region in the latter portion. Scale bars represent 2 μm in the larger view and 300 nm in the zoomed view. Time indicated in min:sec. (AVI 5790 kb)

TIRF-SIM reveals filament partitioning in fly cells.

Drosophila cell expressing GFP-squash sampled every 3 seconds with TIRF-SIM. Magenta box in larger image indicates zoomed in region. Playback rate is 15 frames per second. Time in min:sec. Scale bars represent 3 μm larger view and 300 μm in zoomed view. (AVI 6956 kb)

NM2A and NM2B localization in primary MEF.

EGFP-NM2A MEF (cyan) expressing NM2B-halo-JF549 (magenta) sampled every 5 seconds with Zeiss Airyscan. Playback rate is 25 frames per second. Time in min:sec. Scale bar represents 5 μM. (AVI 10352 kb)

Partitioning does not involve axial movement of filaments from out of the imaging plane.

Z-stacks were acquired of EGFP-NM2A MEFs with Airyscan every 10 s with 0.25 μm steps. Max-projection (left), overhead view of 3D rendering (middle), and orthogonal view of 3D rendering (right). Playback rate is 12 frames per second. Time in min:sec. Scale bar represents 300 nm. (AVI 4516 kb)

Partitioning occurs at different EGFP-NM2A intensities.

EGFP-NM2A MEFs sampled every 5 seconds with 2D-SIM. One example each of a dim, medium, and bright partitioning event is shown. Data sets were truncated to just show the partitioning event but nucleation and growth was observed for each. Calibration bar indicates pixel intensities in the upper right corner. Scale bar represents 300 nm. Time in min:sec. (AVI 35162 kb)

Lamellar actin is dense and dynamic.

C57B6 MEF expressing EGFP-FTractin sampled every 5 seconds with TIRF-SIM. Playback rate is 15 frames per second. Time in min:sec. Scale bar represents 3 μm. (AVI 6496 kb)

Actin movements correlate with partitioning – Airyscan.

EGFP-NM2A MEF expressing mApple-FTractin sampled every 10 s with Airyscan. EGFP-NM2A pseudocolored in fire LUT (left) or magenta (right) and mApple-FTractin shown in grey scale (middle, right). Playback rate is 10 frames per second. Time in min:sec. Scale bar represents 300 nm. (AVI 1190 kb)

Actin movements correlate with partitioning – Airyscan2.

EGFP-NM2A MEF expressing mApple-FTractin sampled every 5 s with Airyscan. EGFP-NM2A pseudocolored in fire LUT (left) or magenta (right) and mApple-FTractin shown in grey scale (middle, right). Playback rate is 15 frames per second. Time in min:sec. Scale bar represents 300 nm. (AVI 9313 kb)

Actin movements correlate with partitioning – TIRF-SIM.

EGFP-NM2A MEF (green) expressing tdTomato-FTractin (red) sampled every 5 s with TIRF-SIM. Playback rate is 12 frames per second. Time in min:sec. Scale bar represents 300 nm. (AVI 8138 kb)

Actin dynamics are stalled with SMIFH2 and CK666.

C57B6 MEF expressing Halo-JF594-FTractin sampled every 10 s with Airyscan. SMIFH2 (5 μm) and 5 μM CK666 (5 μm) added when indicated. Playback rate is 24 frames per second. Time in min:sec. Scale bar represents 3 μm. (AVI 19568 kb)

Example of maxima tracking.

EGFP-NM2A MEF sampled every 5 seconds with TIRF-SIM. Playback rate is 30 frames per second. Time in min:sec. Scale bar represents 300 nm. (AVI 2206 kb)

Rock inhibition induces rapid and dramatic redistribution of EGFP-NM2A.

EGFP-NM2A MEF (cyan) expressing mApple-FTractin (magenta) sampled every 30 s with Nikon A1R confocal. Y27632 (10 μM) added when indicated. Playback rate is 30 frames per second. Time in min:sec. Scale bar represents 5 μm. (AVI 19297 kb)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Beach, J., Bruun, K., Shao, L. et al. Actin dynamics and competition for myosin monomer govern the sequential amplification of myosin filaments. Nat Cell Biol 19, 85–93 (2017). https://doi.org/10.1038/ncb3463

Download citation

Further reading

Search

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing